Citation Recommendation Algorithms Based On Multi-dimensional Fusion

With the rapid development and widespread popularity of the Internet,the level of scientific research in various fields of the society is constantly improving.Science and technology are developing rapidly,and the scientific literature is an important product.In this process,the number of scientific literature is increasing.How to obtain the literature required by scholars from a large database of documents is an urgent problem for researchers.In the face of such a problem,the recommendation system came into being and became an indispensable part of the era of big data.In recent years,citation recommendation has attracted more and more researchers in the academic field,and has become a research hotspot in the academic field and the commercial field.At present,the related methods of citation recommendation consider the characteristics of citations to be single,mostly from a certain attribute or index.However,in reality,the citation relationship is complex,and the evaluation of single identification indicators or attributes is prone to bias,lacking the overall analysis of citations.Therefore,the comprehensive consideration of dimensions has become an idea for scholars to study citations.In this paper,the related research work is carried out for citation recommendation,and a citation recommendation algorithm with multi-dimensional fusion is proposed.The main research work is as follows:1.Based on the network diagram of literature citation,this paper considers the three issues of literature difference,time and low-cited of high quality literature,and proposes the importance analysis of the literature by New-PR algorithm.The experimental results show that the performance of the literature recommended by the New-PR algorithm is significantly better than that of the original algorithm.2.This paper uses feature words and pattern recognition methods to accurately locate sentences that describe innovation points,and based on candidate sentences and key phrases,establishes an evaluation model for scientific literature innovation.The experimental results show that the scientific literature innovation evaluation model proposed in this paper has reasonableness and effectiveness in literature evaluation and recommendation.3.Based on the restart random walk algorithm and word2 vec model,this paper improves and optimizes the probability transfer matrix of the original algorithm to better adapt to the analysis and calculation of literature relevance.The experimental results show that the improved algorithm is superior to the original algorithm in the recall rate and NDCGindex performance.This paper proposes a multi-dimensional fusion algorithm that effectively combines the document importance dimension,the literature relevance dimension,the literature innovation dimension,the document activity degree and the author dimension.The experimental results show that the five dimensions have positive influence on the citation recommendation,and the citation recommendation algorithm proposed in this paper is more reasonable than other algorithms.Because of considering multiple dimensions,the candidate collection has a more reasonable score value.Therefore,the final recommendation effect is also more ideal.